|
|
Post by BioLogIn on Feb 10, 2008 16:28:50 GMT -5
biologin.guiltygear.ru/warlocks_league.php as of 2019, this script is no longer useful or available A simple PHP script to generate parings for league. Accepts an (even) number of player names (there must be more then 5 and less then 99 names). Generates either completely random or ELO-based league parings. ELO-based parings are done this way: fetch ELO from RB server (takes a while, so please be patient), sort players by ELO, split them into 5 groups, and try to assign 5 opponents (one randomly chosen from each group) to each of players. Please note that if "randomize parings" option was checked, no name validation will be done. But if parings are to be done via ELO, all names will be validated (to fetch ELO) and all misspelled names will be discarded. biologin.guiltygear.ru/warlocks_teams.php as of 2019, this script is no longer useful or available A very simple PHP script to generate random teams for team melee. Accepts 4 or 6 names (strings), then generates random teams. For 6 names both 2v2v2 and 3v3 teams will be generated. No names validation or anything. === Feedback is welcome. |
|
|
|
Post by Rycchus on Feb 10, 2008 17:03:34 GMT -5
Neat.
For the league version, a "number of matches" option might be nice, so that it's not constricted to five.
|
|
|
|
Post by BioLogIn on Feb 12, 2008 4:57:34 GMT -5
Neat. For the league version, a "number of matches" option might be nice, so that it's not constricted to five. Thanks I guess I'll add this option later ) |
|
|
|
Post by vermont on Feb 12, 2008 8:00:20 GMT -5
BioLogIn, I especially like the code that makes it so you never have to play ExDeath or Slarty.  |
|
|
|
Post by BioLogIn on Feb 12, 2008 8:15:11 GMT -5
Vermont Ok, I'll try to add this in the next version === Newer version of league script is available. Added "Number of matches" feature. Added some ELO stats in output. Improved algorithm that divides players into groups. === As usual, feedback is welcome ) |
|
|
|
Post by Rycchus on Feb 12, 2008 10:42:27 GMT -5
You should either say in the notes that you can't have more games than opponents, or have an "allow repeats" checkbox.
I like the "opponent average ELO" thing.
First time I've tested the non-random pairing, but it still seems pretty random to me - the average opponents' ELO ranges from 1713 to 1905. And 1905 was got by the second-lowest player, so I'm assuming you didn't weight it. Maybe your system is only effective for a large number of contestants? (I had 12, 5 games each)
|
|
|
|
Post by BioLogIn on Feb 12, 2008 11:00:30 GMT -5
You should either say in the notes that you can't have more games than opponents, or have an "allow repeats" checkbox. I've written "Please note that amount of participants and amount of games for each player are related". IMHO that's clear enough for everobody who knows how our league paring goes. Do you think it needs further explanation. I like the "opponent average ELO" thing. First time I've tested the non-random pairing, but it still seems pretty random to me - the average opponents' ELO ranges from 1713 to 1905. And 1905 was got by the second-lowest player, so I'm assuming you didn't weight it. Maybe your system is only effective for a large number of contestants? (I had 12, 5 games each) Yeah, that's what I wanted to bring attention to. We have relatively small number of players, and that gives us HUGE difference in rating even in one "rating group". For example, if we test this script on current league participants with their current ELO, in the "top" group we will have 5 players with ELO from 2038 to 1884, and in the bottom we wil have spread from 1561 to 1717. Since we're picking players from ELO-groups randomly, it is still possible to have a differrence of 50 avg. opp. ELO points. So my personal opinion is that for such small leagues this system (ELO pools seeding) isn't big improvement over random system. It isn't worse either, so it is up to league org to use it or not to use it. It is possible to implement LS-method that will build parings with minimal avg. opp. ELO difference between any two players, but - these pairings will not be random at all, and that sucks IMO - I don't want to implement it at all =) |
|
|
|
Post by Rycchus on Feb 12, 2008 11:23:31 GMT -5
A difference of 50 isn't quite so bad - it was the difference of 200 that was drastic.
Well, you actually put
Please note that amount of participants and amount of games for each player are related, and their multiplication should be an even number.
which makes it sound like the relation you're talking about is an even product. Maybe I read too quickly. and note that would clear the ambiguity, or putting something about there being no repeats - it is after all possible to run a league with repeats in.
|
|
|
|
Post by BioLogIn on Mar 3, 2008 9:08:42 GMT -5
As for now, script is unable to pair an odd number of participants correctly. I'll try to fix this soon.
|
|
|
|
Post by BioLogIn on Mar 31, 2008 10:06:33 GMT -5
Script now should work correctly with any number of participants.
|
|
